diff --git a/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/infer/concat_infer.c b/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/infer/concat_infer.c
index 81fda9c0b65..e6b6798067c 100644
--- a/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/infer/concat_infer.c
+++ b/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/infer/concat_infer.c
@@ -65,11 +65,7 @@ int ConcatInferShape(const TensorC *const *inputs, size_t inputs_size, TensorC *
   for (size_t i = 1; i < inputs_size; ++i) {
     size_t input_i_shape_size = inputs[i]->shape_size_;
     if (input_i_shape_size != input0_shape_size) {
-      if (input_i_shape_size != 0) {
-        return NNACL_PARAM_INVALID;
-      } else {
-        continue;
-      }
+      return NNACL_PARAM_INVALID;
     }
     int shape_tmp[MAX_SHAPE_SIZE] = {0};
     size_t shape_tmp_size = 0;
diff --git a/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/infer/crop_infer.c b/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/infer/crop_infer.c
index 2c0393c0fe6..5f2ca237df7 100644
--- a/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/infer/crop_infer.c
+++ b/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/infer/crop_infer.c
@@ -16,6 +16,7 @@
 
 #include "nnacl/infer/crop_infer.h"
 #include "nnacl/infer/infer_register.h"
+#include "nnacl/crop_parameter.h"
 
 int CropInferShape(const TensorC *const *inputs, size_t inputs_size, TensorC **outputs, size_t outputs_size,
                    OpParameter *parameter) {
@@ -24,6 +25,13 @@ int CropInferShape(const TensorC *const *inputs, size_t inputs_size, TensorC **o
     return check_ret;
   }
 
+  size_t input_shape_size = inputs[0]->shape_size_;
+  CropParameter *param = (CropParameter *)parameter;
+  int64_t axis = param->axis_ < 0 ? param->axis_ + (int64_t)input_shape_size : param->axis_;
+  if (axis < 0 || axis >= input_shape_size) {
+    return NNACL_ERR;
+  }
+
   SetDataTypeFormat(outputs[0], inputs[0]);
   if (!InferFlag(inputs, inputs_size)) {
     return NNACL_INFER_INVALID;
diff --git a/mindspore/lite/src/runtime/kernel/arm/base/crop_base.cc b/mindspore/lite/src/runtime/kernel/arm/base/crop_base.cc
index 06954fdfcb2..f63280bd7b6 100644
--- a/mindspore/lite/src/runtime/kernel/arm/base/crop_base.cc
+++ b/mindspore/lite/src/runtime/kernel/arm/base/crop_base.cc
@@ -49,7 +49,7 @@ int CropBaseCPUKernel::ReSize() {
 int CropBaseCPUKernel::PadOffset(int input_dim, CropParameter *crop_para) const {
   auto axis = crop_para->axis_;
   auto offsets_size = crop_para->offset_size_;
-  MS_ASSERT(axis <= input_dim);
+  MS_CHECK_TRUE_MSG(axis < input_dim, RET_ERROR, "The axis is invalid.");
   if (offsets_size > 1) {
     MS_CHECK_TRUE_MSG(axis + offsets_size == input_dim, RET_ERROR, "The axis and offsets is invalid");
   }
diff --git a/mindspore/lite/src/runtime/kernel/arm/fp32/gatherNd_fp32.cc b/mindspore/lite/src/runtime/kernel/arm/fp32/gatherNd_fp32.cc
index f0239e76574..8163632f50a 100644
--- a/mindspore/lite/src/runtime/kernel/arm/fp32/gatherNd_fp32.cc
+++ b/mindspore/lite/src/runtime/kernel/arm/fp32/gatherNd_fp32.cc
@@ -105,7 +105,7 @@ int GatherNdCPUKernel::InitOffset() {
   (void)memset(in_offset_, 0, count_ * sizeof(int));
   for (int j = 0; j < count_; ++j) {
     for (int k = 0; k < idx_lastshape; ++k) {
-      if (indices_ptr[j * idx_stride + k] < in_shape[k]) {
+      if (indices_ptr[j * idx_stride + k] >= 0 && indices_ptr[j * idx_stride + k] < in_shape[k]) {
         in_offset_[j] += indices_ptr[j * idx_stride + k] * in_stride.at(k);
       } else {
         MS_LOG(ERROR) << name() << " indices value invalid!";
diff --git a/mindspore/lite/src/runtime/kernel/arm/fp32/matmul_fp32_base.cc b/mindspore/lite/src/runtime/kernel/arm/fp32/matmul_fp32_base.cc
index 8f131e80bb4..de8492ca9df 100644
--- a/mindspore/lite/src/runtime/kernel/arm/fp32/matmul_fp32_base.cc
+++ b/mindspore/lite/src/runtime/kernel/arm/fp32/matmul_fp32_base.cc
@@ -312,7 +312,11 @@ int MatmulFp32BaseCPUKernel::Prepare() {
   CHECK_LESS_RETURN(in_tensors_.size(), C2NUM);
   CHECK_LESS_RETURN(out_tensors_.size(), 1);
   init_global_variable();
+  MS_CHECK_INT_MUL_NOT_OVERFLOW(a_batch_, params_->row_align_, RET_ERROR);
+  MS_CHECK_INT_MUL_NOT_OVERFLOW(a_batch_ * params_->row_align_, params_->deep_, RET_ERROR);
   matrix_a_pack_size_ = a_batch_ * params_->row_align_ * params_->deep_;
+  MS_CHECK_INT_MUL_NOT_OVERFLOW(a_batch_, params_->col_align_, RET_ERROR);
+  MS_CHECK_INT_MUL_NOT_OVERFLOW(a_batch_ * params_->col_align_, params_->deep_, RET_ERROR);
   matrix_b_pack_size_ = b_batch_ * params_->col_align_ * params_->deep_;
   if (matrix_a_pack_size_ < 0 || matrix_b_pack_size_ < 0) {
     MS_LOG(ERROR) << "Matrix pack size is negative "
diff --git a/mindspore/lite/src/runtime/kernel/arm/fp32/reverse_fp32.cc b/mindspore/lite/src/runtime/kernel/arm/fp32/reverse_fp32.cc
index 28f2f1163fb..f4e2e958003 100644
--- a/mindspore/lite/src/runtime/kernel/arm/fp32/reverse_fp32.cc
+++ b/mindspore/lite/src/runtime/kernel/arm/fp32/reverse_fp32.cc
@@ -37,7 +37,11 @@ int ReverseCPUKernel::Stride(int index) {
 
 int ReverseCPUKernel::ReSize() {
   // trans negative to positive axis
-  UpdateAxisInfo();
+  auto ret = UpdateAxisInfo();
+  if (ret != RET_OK) {
+    MS_LOG(ERROR) << "Get axis failed.";
+    return ret;
+  }
 
   data_size_ = in_tensors_.at(0)->ElementsNum();
   thread_sz_count_ = MSMIN(op_parameter_->thread_num_, data_size_);
@@ -146,14 +150,19 @@ int ReverseCPUKernel::Run() {
   return RET_OK;
 }
 
-void ReverseCPUKernel::UpdateAxisInfo() {
+int ReverseCPUKernel::UpdateAxisInfo() {
   auto reverse_param = reinterpret_cast<ReverseParameter *>(op_parameter_);
   int in_shape_len = static_cast<int>(in_tensors_.front()->shape().size());
   for (int i = 0; i < reverse_param->num_axis_; ++i) {
     if (reverse_param->axis_[i] < 0) {
       reverse_param->axis_[i] += in_shape_len;
     }
+    if (reverse_param->axis_[i] < 0 || reverse_param->axis_[i] >= in_shape_len) {
+      MS_LOG(ERROR) << "Invalid axis.";
+      return RET_ERROR;
+    }
   }
+  return RET_OK;
 }
 
 REG_KERNEL(kCPU, kNumberTypeFloat32, PrimitiveType_ReverseV2, LiteKernelCreator<ReverseCPUKernel>)
diff --git a/mindspore/lite/src/runtime/kernel/arm/fp32/reverse_fp32.h b/mindspore/lite/src/runtime/kernel/arm/fp32/reverse_fp32.h
index 08c81379106..571493ac5a3 100644
--- a/mindspore/lite/src/runtime/kernel/arm/fp32/reverse_fp32.h
+++ b/mindspore/lite/src/runtime/kernel/arm/fp32/reverse_fp32.h
@@ -42,7 +42,7 @@ class ReverseCPUKernel : public InnerKernel {
   int DoReverse(int task_id);
 
  private:
-  void UpdateAxisInfo();
+  int UpdateAxisInfo();
   int thread_sz_count_ = 0;
   int thread_sz_stride_ = 0;
   int data_size_ = 0;
diff --git a/mindspore/lite/src/runtime/kernel/arm/int8/softmax_int8.cc b/mindspore/lite/src/runtime/kernel/arm/int8/softmax_int8.cc
index 78e51c99d58..0ca2691230d 100644
--- a/mindspore/lite/src/runtime/kernel/arm/int8/softmax_int8.cc
+++ b/mindspore/lite/src/runtime/kernel/arm/int8/softmax_int8.cc
@@ -51,6 +51,7 @@ int SoftmaxInt8CPUKernel::Prepare() {
   MS_ASSERT(input_tensor != nullptr);
 
   auto in_quant_args = input_tensor->quant_params();
+  CHECK_LESS_RETURN(in_quant_args.size(), 1);
   quant_param_->in_quant_args_.scale_ = static_cast<float>(in_quant_args.front().scale);
   quant_param_->in_quant_args_.zp_ = -in_quant_args.front().zeroPoint;
 
@@ -58,6 +59,7 @@ int SoftmaxInt8CPUKernel::Prepare() {
   MS_ASSERT(out_tensor != nullptr);
 
   auto out_quant_args = out_tensor->quant_params();
+  CHECK_LESS_RETURN(out_quant_args.size(), 1);
   quant_param_->out_quant_arg_.scale_ = static_cast<float>(out_quant_args.front().scale);
   quant_param_->out_quant_arg_.zp_ = -out_quant_args.front().zeroPoint;
   quant_param_->output_activation_min_ = std::numeric_limits<int8_t>::min();